I've had some version of this conversation about forty times this year. A business owner pulls me aside, lowers their voice slightly, and asks: "So honestly, how many people am I going to need in two years?" The honest answer is probably the same number. But doing different things.
What the Data Says
PwC's 2025 Global AI Jobs Barometer analysed close to a billion job postings across six continents. The headline: jobs requiring AI skills command a 56% wage premium over similar roles that don't, up from 25% the year before. Workers with AI capabilities are becoming more valuable, not less.
The Yale Budget Lab has been tracking AI's actual labour market impact through monthly survey data. Their conclusion: "The broader labour market has not experienced a discernible disruption since ChatGPT's release." Employment measures, unemployment rates, occupational mix. All within historical ranges.
CNBC reported in September 2025 that AI's workforce impact is "small, but not zero." The Brookings Institution found that employers are more likely to retrain workers than lay them off. Meanwhile, productivity growth nearly quadrupled in AI-exposed industries between 2018 and 2024.
The people did not disappear. They got more done.
What Is Actually Changing Inside Teams
The studies converge on something that matters more than headcount: the composition of work is shifting. Fast. PwC found that skills employers seek are changing 66% faster in AI-exposed occupations than in others.
In a 10-to-50 person company, that looks like this. The routine task layers are compressing. Data entry. First-draft writing. Report generation. Basic research. The time they take is shrinking, sometimes by half or more. A person who spent four hours a day on data consolidation now spends ninety minutes. The other two and a half hours need to go somewhere.
The work that's expanding sits at the other end. Client relationships. Professional judgment. Exception handling. The phone call where a frustrated customer needs someone who actually listens. Trust-based work. Contextual work.
The Safety Question Nobody Is Asking
I don't mean data safety. I mean psychological safety. The teams that will get the most from AI are the ones where people feel comfortable saying "I tried this with AI and it didn't work" without worrying about looking incompetent. Where a junior employee can say "I used Claude to draft this analysis and then revised it" without their manager questioning their value.
That culture doesn't build itself. It requires leadership that is visibly experimenting too, not just mandating adoption from above.
What to Do With This Information
If you lead a small or mid-sized team: you probably don't need fewer people in the next eighteen months. You almost certainly need your people doing different work.
Start by auditing where your team's hours actually go. Not where you think they go. Track it for two weeks. You'll find pockets of repetitive work that everyone has just accepted. Those pockets are where AI creates the most immediate value, not by eliminating the person, but by freeing them up for work that actually requires their experience.
AI is not coming for your team. But it is coming for the way your team spends its time. The companies that recognise the difference will have a real advantage in hiring, retention, and output.